Docs
Get rtrvr running.
Start in Chrome, run in Cloud, or call the agent from your product. Pick a path and finish the first job.
Search docs⌘ KStart here
Choose how you want to work.
Pick the environment closest to the job. You can add another one when you need it.
Start in Chrome.
Work in the browser where you are already signed in. Ask for a table, a filled form, or a completed update.
Give your product a browser.
Call the agent from your app, CLI, SDK, or MCP client. Send the job and starting URLs. Get structured output back.
Run beyond your laptop.
Run a task in a separate Cloud browser. Choose a subscription when it needs to run on a schedule.
Give your agent the docs.
Install the rtrvr skill in Claude Code or Codex. Your agent loads these docs whenever you ask about rtrvr, from writing tasks to calling the API.
mkdir -p ~/.claude/skills && curl -fsSL https://rtrvr.ai/skill.zip -o /tmp/rtrvr.zip && unzip -o /tmp/rtrvr.zip -d ~/.claude/skills
Using Codex? Unzip into ~/.codex/skills instead.
The first prompt
Say what you need done.
Give rtrvr the goal and the page or starting URL. It plans the route, does the work, and returns what you asked for.
Compare the current plans on this page. Return a table with plan, price, billing period, and source links.
Side by side
One task. Three outcomes.
Browser Use gets stuck repeating steps. Firecrawl dumps scraped URLs and hallucinates the answer. rtrvr plans the research, traverses the right sources, and returns a cited answer.
Read the full comparison